I am running hadoop-2.4.0 cluster. Each datanode has 10 disks, directories for 10 disks are specified in dfs.datanode.data.dir.
A few days ago, I modified dfs.datanode.data.dir of a datanode () to reduce disks. so two disks were excluded from dfs.datanode.data.dir, after the datanode was restarted, I expected that the namenode would update block locations. In other words, I thought the namenode should remove from block locations associated with blocks which were stored on excluded disks, but the namenode didnt update the block locations...
In my understanding, datanode send a block report to the namenode when datanode start so the namenode should update block locations immediately.
Is a bug? Could anyone please explain?
